The objective of this project is to improve the accuracy and precision of dose targeting in image-guided radiation treatment by accurately and automatically registering deformable tissues in diagnostic and treatment planning images. Recent advances in medical imaging and radiation therapy have the potential to improve patient care by noninvasively identifying the location and extent of cancer and by allowing physicians to escalate and target radiation dose to cancerous lesions while sparing surrounding healthy tissues. To compensate for significant changes that occur between diagnostic and treatment phases due to imaging requirements such as endorectal probes, patient position differences, weight change, and other factors, a new computational tool for deformable image registration will be developed and validated. Given a segmented reference image and an image obtained during treatment, the algorithm will generate a 3D representation of the tissues, estimate tissue displacements and deformations using a finite element model, account for uncertainty due to unknown model parameters, and output a mapping between the images. Low computation time is critical for clinical viability. The new method will be applied to register prostate CT or MRI/MRSI reference images to treatment images obtained using Megavoltage Cone-Beam GT (MV CBCT), a 3D imaging modality being developed to image the patient on the treatment table for Intensity-Modulated Radiation Therapy (IMRT). Validation of the method will be performed using bone outlines and by measuring registration errors for marker seeds that have been implanted inside the prostate in a subset of patients. The new software tool, targeted at prostate cancer treatment, aims to significantly improve patient care by enabling clinicians to improve conformality of dose to tissue types during radiation therapy. These improvements could significantly enhance public health by lowering recovery times, recurrence rates, and treatment costs for cancer patients. [unreadable] [unreadable] [unreadable]